Research Degree - Providing Personalized Information Based on Individual Interests and Preferences
Communication and Computing Research Centre
2010 - ongoing
Research Degree Project
The main focus of this research is to devise a framework of a Personalised Search Engine (PSE) to re-rank the URLs of each search result returned by a non-personalised search engine. The main goal of a personalised search engine is to tailor the search results to the needs of individual users and provide them with personalised and adapted information based on their interests and preferences. In this study, the gaps in the personalisation of the search results returned by a standard search engine in response to the searcher's query are identified, the searcher's individual interests and preferences are investigated, and a framework of PSE for the re-ranking processes is proposed based on individual user's interests and preferences.
The research objectives are twofold: (1) to use the search logs or browsing histories to determine the Web searchers' interests; (2) to gain an insight into how to match these interests with the users' needs and preferences to determine the relevance of each document.
For the development of the PSE framework, a novel approach which combines the Vector Space Model (VSM) as a weighting schema for the keyword frequency, with the Profile Ontology (PO) for measuring the relationship between keywords and their corresponding links, is proposed to enhance the degree of personalisation. In this approach, the key research question being explored is whether the hybridisation of the two techniques - the VSM and the PO - can provide a more effective and personalised ranking of the search results than the standard web search engines. The hybridisation is enhanced by incorporating in the evaluation, the dwell time based model of personal relevance to distinguish a highly relevant document from a moderately relevant document.
For the pilot and the main studies of Web interactions, a first cycle of search logs were collected from 50 respondents and a second cycle were later collected from 48 participants who all volunteered to be recruited. Their search logs were gathered for the ranking algorithms of the personalised search engine. These search logs from consenting individual users represented their search browsing histories and were used to build the users’ profiles and construct the POs to establish the relations of various keywords and links for each input keyword. The importance of each visited document is established based on the weight of the entered keywords' frequency. The search results of the PSE is re-ranked based on the weight in each individual document and in all documents related to the keywords.
This proposed framework is a complete study of the personalisation process which provides quality search results ranked in their order of relevance based on individual searcher's interests and preferences. It is also a ranking process gap identifier. Its inductive research approach adopted both quantitative and qualitative techniques for the collection of data and its analysis. Its output is a data mining model especially applicable to researchers who see the benefits of personalisation. While other studies provide personalisation of recommenders' systems, this study focuses on the personalisation of the search results for researchers to address an area of Information Retrieval (IR). The project is the first of its kind to add the dwell time to establish the degree of relevance for a document.
- Dr Martin Beer(Director of Study)
- Dr Elizabeth Uruchurtu (Second Supervisor)
Safiya Al Sharji - Research Degree Student